基于FRFT自相似参数估计的异常流量检测方法

针对传统异常流量检测方法检测精度较低, Hurst指数估计受估计序列尺度的影响, 提出了基于分数阶傅里叶变换（FRFT）估计Hurst指数的方法。在此基础上, 实现了基于Hurst指数变化的异常检测, 有效解决了方法实现过程中FRFT最佳估计的分数阶阶数选择及Hurst参数求解的关键问题。实验表明, 基于FRFT的估计不受序列非平稳性影响, 对Hurst指数估计具有较高的估计精度, 并且可以准确地检测网络异常。

Anomaly traffic detection using Hurst estimation based on FRFT

Since traditional abnormal detection methods have poor performance, and Hurst parameter estimation was affected by non-stationary traffic, this paper designed the FRFT estimation method. On that basis, it implemented the abnormal detection method based on Hurst parameter variation. In addition, it resolved key issues of choice of suitable fractional order and calculation of Hurst. The experimental results show that FRFT estimation method is not affect by non-stationary traffic, which has better performance on Hurst estimation, and FRFT detection method identify abnormal traffic accurately.